Age | Commit message (Collapse) | Author |
|
-ITE Simplification
-- Moved the utilities in src/theory/ite_simplifier.{h,cpp} to ite_utilities.
-- Separated simpWithCare from simpITE.
-- Disabled ite simplification on repeat simplification by default. Currently, ite simplification cannot help unless we internally make new constant leaf ites equal to constants.
-- simplifyWithCare() is now only run on QF_AUFBV by default. Speeds up nec benchmarks dramatically.
-- Added a new compress ites pass that is only run on QF_LIA by default. This targets the perverse structure of ites generated during ite simplification on nec benchmarks.
-- After ite simplification, if the ite simplifier was used many times and the NodeManager's node pool is large enough, this garbage collects: zombies from the NodeManager repeatedly, the ite simplification caches, and the theory rewrite caches.
- TheoryEngine
-- Added TheoryEngine::donePPSimpITE() which orchestrates a number of ite simplifications above.
-- Switched UnconstrainedSimplifier to a pointer.
- RemoveITEs
-- Added a heuristic for checking whether or not a node contains term ites and if not, not bothering to invoke the rest of RemoveITE::run(). This safely changes the type of the cache used on misses of run. This cache can be cleared in the future. Currently disabled pending additional testing.
- TypeChecker
-- added a neverIsConst() rule to the typechecker. Operators that cannot be used in constructing constant expressions by computeIsConst() can now avoid caching on Node::isConst() calls.
- Theory Bool Rewriter
-- Added additional simplifications for boolean ites.
Minor Changes:
- TheoryModel
-- Removed vestigial copy of the ITESimplifier.
- AttributeManager
-- Fixed a garbage collection bug when deleting the node table caused the NodeManager to reclaimZombies() which caused memory corruption by deleting from the attributeManager.
- TypeChecker
-- added a neverIsConst() rule to the typechecker. Operators that cannot be used in constructing constant expressions by computeIsConst() can now avoid caching on Node::isConst() calls.
-NodeManager
-- Added additional functions for reclaiming zombies.
-- Exposed the size of the node pool for heuristics that worry about memory consumption.
- NaryBuilder
-- Added convenience classes for constructing associative and commutative n-ary operators.
-- Added a pass that turns associative and commutative n-ary operators into binary operators. (Mostly for printing expressions for strict parsers.)
|
|
being able to list attributes. Additionally, added debugging hooks to node manager and attribute manager.
|
|
|
|
|
|
|
|
|
|
Also support array-store-all for Boolean terms (related to abstract values, since that's the only way for the user to include an array-store-all in an assertion).
(this commit was certified error- and warning-free by the test-and-commit script.)
|
|
Also some fixes to parametric datatypes I found, and fixes for a handful of bugs, including some observed with --check-models --incremental on together.
(this commit was certified error- and warning-free by the test-and-commit script.)
|
|
just the header comments at the top, though. Don't update to this rev if
you don't have time for a complete rebuild, and exclude this rev if you
want to see what's new across a range of commits.
(this commit was certified error- and warning-free by the test-and-commit script.)
|
|
make it unambiguous for case-insensitive filesystems like on Mac. Fixes Mac builds
|
|
The external interface (e.g., what's answered by ExprManager::getStatistics() and SmtEngine::getStatistics()) is a snapshot of the current statistics (rather than a reference to the actual StatisticsRegistry).
The StatisticsRegistry is now internal-only. However, it's built as a convenience library so that the parser and driver can use it too (by re-linking against it).
This is part of the ongoing effort to clean up the public interface.
(this commit was certified error- and warning-free by the test-and-commit script.)
|
|
meeting last week. The SmtEngine now subscribes to NodeManager events,
does appropriate dumping of variable declarations, and notifies the Model
class.
The way to create a skolem is now:
nodeManager->mkSkolem("myvar_$$", TypeNode, "is a variable created by the theory of Foo")
The first argument is the name of the skolem, and the (optional) "$$" is a
placeholder for the node id (to get a unique name). Without a "$$", a "_$$"
is automatically appended to the given name.
The second argument is the type.
The (optional, but recommended) third argument is a comment, used by the
dump infrastructure to indicate what the variable is for / who owns it.
An optional fourth argument (not shown) allows you to specify flags that
control the behavior (e.g., don't do notification, and/or don't make a
unique name). Look at the documentation for details on these.
In particular, the above means you can't just do a mkSkolem(boolType) for
example---you have to specify a name and (hopefully also,
but it's optional) a comment. This leads to easier debugging than the
anonymous skolems before, since we'll be able to track where the skolems
came from.
Much of the Model and Dump stuff, as well as some Command stuff, is cleaned up
by this commit. Some remains to be cleaned up.
(this commit was certified error- and warning-free by the test-and-commit script.)
|
|
1. changes the way options are declared (see http://church.cims.nyu.edu/wiki/Options)
2. moves module-specific options enumerations (SimplificationMode, DecisionMode, ArithUnateLemmaMode, etc.) to their own header files, also they are no longer inside the Options:: class namespace.
3. includes many SMT-LIBv2 compliance fixes, especially to (set-option..) and (get-option..)
The biggest syntactical changes (outside of adding new options) you'll notice are in accessing and setting options:
* to access an option, write (e.g.) options::unconstrainedSimp() instead of Options::current()->unconstrainedSimp.
* to determine if an option value was set by the user, check (e.g.) options::unconstrainedSimp.wasSetByUser().
* ensure that you have the option available (you have to #include the right module's options.h file, e.g. #include "theory/uf/options.h" for UF options)
*** this point is important. If you access an option and it tells you the option doesn't exist, you aren't #including the appropriate options.h header file ***
Note that if you want an option to be directly set (i.e., other than via command-line parsing or SmtEngine::setOption()), you need to mark the option :read-write in its options file (otherwise it's read-only), and you then write (e.g.) options::unconstrainedSimp.set(true).
Adding new options is incredibly simple for primitive types (int, unsigned, bool, string, double). For option settings that you need to turn into a member of an enumerated type, you write a custom "handler" for the option---this is no additional work than it was before, and there are many examples to copy from (a good one is stringToSimplificationMode() in src/smt/options_handlers.h).
Benefits of the new options system include:
1. changes to options declarations don't require a full-source rebuild (you only have to rebuild those sources that depend on the set of options that changed).
2. lots of sanity checks (that the same option isn't declared twice, that option values are in range for their type, that all options are documented properly, etc.)
3. consistency: Boolean-valued option --foo gets a --no-foo automatically, documentation is generated consistently, the option-parsing matches the documented option name, etc.
4. setting options programmatically via SmtEngine::setOption() is enabled, and behaves the same as command-line equivalents (including checking the value is in range, etc.)
5. the notion of options being "set by the user" is now primitive; you can use (e.g.) options::unconstrainedSimp.wasSetByUser() instead of having to use (and maintain) a separate Boolean option for the purpose
I've taken lots of care not to break anything. Hopefully, I've succeeded in that.
|
|
See below for details.
* Fix the "assert" name-collision bug (resolves bug #364).
Our identifiers should never be named "assert", as that's a preprocessor
definition in <assert.h>, which is often #included indirectly (so simply
having a policy of not including <assert.h> isn't good enough---one of
our dependences might include it). It was once the case that we didn't
have anything named "assert", but "assert()" has now crept back in.
Instead, name things "assertFoo()" or similar. Thanks to Tim for the
report.
To fix this, I've changed some of Dejan's circuit-propagator code from
"assert()" to "assertTrue()". Ditto for Andy's explanation manager.
Guys, if you prefer a different name in your code, please change it.
* Fix the incorrect parsing of lets in SMT-LIBv2 parser (resolves bug #365).
Inner lets now shadow outer lets (previously, they incorrectly gave an
error). Additionally, while looking at this, I found that a sequential let
was implemented rather than a parallel let. This is now fixed. Thanks to
Liana for the report.
* Remove ANTLR parser generation warnings in CVC parser (resolves bug #314).
* There were a lot of Debug lines in bitvectors that had embedded toString()
calls. This wasted a LOT of time in debug builds for BV benchmarks
(like in "make regress"). Added if(Debug.isOn(...)) guards; much faster
now.
* Support for building public-facing interface documentation only (as opposed
to all internals documentation). Now "make doc" does the public-facing and
"make doc-internals" does documentation of everything. (Along with changes
to the nightly build script---which will now build and publish both types
of Doxygen documentation---this resolves bug #359).
* Fix the lambda typechecking bug (resolves bug #322). Thanks to Andy for the
report (a long long time ago--sorry).
* The default output language for all streams is now based on the current set
of Options (if there is one). This has been a constant annoyance, especially
when stringstreams are used to construct output. However, it doesn't work
for calls from outside the library, so it's mainly an annoyance-fixer for
CVC4 library code itself.
* Add some CVC4_UNUSED markers to local variables in theory_arith.cpp that
are used only in assertions-enabled builds (and thus give warnings in
production builds). This was briefly discussed at the meeting this week.
|
|
support incrementality.
Some clean-up work will likely follow, but the CNF/Minisat stuff should be
left pretty much untouched.
Expected performance change negligible; slightly better on memory:
http://church.cims.nyu.edu/regress-results/compare_jobs.php?job_id=3705&reference_id=3697&mode=&category=&p=5
Note that there are crashes, but that these are exhibited in the nightly
regression run too!
|
|
SMT-LIBv1 and SMT-LIBv2 input:
In SMT-LIBv1, you specify the "cvc4_logic" benchmark attribute; for instance:
(benchmark actually_a_sat_benchmark_but_looks_like_uf
:logic QF_UF
:cvc4_logic { QF_SAT }
[...]
In SMT-LIBv2, you use a set-info; for instance:
(set-logic QF_UF)
(set-info :cvc4-logic "QF_SAT")
[...]
Right now, the only thing this does is disable the symmetry breaker for
benchmarks like the above ones.
As part of this work, TheoryEngine::setLogic() was removed (the logic field there
wasn't actually used anywhere, its need disappeared when
Theory::setUninterpretedSortOwner() was provided).
Also, Theory::d_uninterpretedSortOwner got a name change to
Theory::s_uninterpretedSortOwner, to highlight that it is static to the Theory
class. This represents a breakage of our separation goals for CVC4, since it
means that two SmtEngines cannot be created separately to solve a QF_AX and
QF_UF problem. A bug report is pending.
|
|
|
|
Datatypes (bug #283) by Chris Conway. Thanks, Chris!
|
|
Dumping infrastructure. Can dump preprocessed queries and clauses. Can
also dump queries (for testing with another solver) to see if any conflicts
are missed, T-propagations are missed, all lemmas are T-valid, etc. For a
full list of options see --dump=help.
CUDD building much cleaner.
Documentation and assertion fixes.
Printer improvements, printing of commands in language-defined way, etc.
Typechecker stuff in expr package now autogenerated, no need to manually
edit the expr package when adding a new theory.
CVC3 compatibility layer (builds as libcompat).
SWIG detection and language binding support (infrastructure).
Support for some Z3 extended commands (like datatypes) in SMT-LIBv2 mode
(when not in compliance mode).
Copyright and file headers regenerated.
|
|
|
|
on some problems---valgrind gave many complaints): the problem was that calloc() (in the Backtracker) wasn't allocating enough space for the type located at the resulting address. Resolves bug #263.
Also, some debugging improvements.
|
|
|
|
Expected performance impact outside of datatypes/CVC parser is
negligible.
* CVC language LAMBDA, functional LET, type LET, precedence fixes,
bitvectors, and arrays, with partial parsing support also for
quantifiers, tuples, subranges, subtypes, and records
* support for complex recursive DATATYPE selectors, e.g.
tree = node(children:ARRAY INT OF tree) | leaf(data:INT)
these are complicated because they have to be left unresolved
at parse time and dealt with in a second pass.
* bugfix for Exprs/Types that occurred when setting them to null
(not Nodes/TypeNodes, just Exprs/Types).
* Cleanup/code review items
|
|
1. Defines a new type "DatatypeType", a type-constant that holds a Datatype, describing an inductive data type.
2. CVC language parser supports datatypes.
3. CVC language printer now functional.
4. Minor other cleanups.
No performance impact is expected outside of datatypes. I'm verifying that that is the case with a cluster job this morning.
|
|
(library-internal-only of course) between Exprs and Nodes, Types and TypeNodes, ExprManagers and NodeManagers.
|
|
* Makes Options an "omnipresent thread-local global" (like the notion
of the "current NodeManager" was already). Options::current() accesses
this structure.
* Removes Options from constructors and data structures everywhere
(this cleans up a lot of things).
* No longer uses StatisticsRegistry statically. An instance of the
registry is created and linked to a NodeManager.
* StatisticsRegistry::current() is similar to Options::current(), but
the pointer is stowed in the NodeManager (rather than stored)
* The static functions of StatisticsRegistry have been left, for backward
compatibility; they now use the "current" statistics registry.
* SmtEngine::getStatisticsRegistry() is a public accessor for the
registry; this is needed by main() to reach in and get the registry,
for flushing statistics at the end.
|
|
from "arrays" branch to trunk
|
|
some macros to make various GCC versions happy
|
|
Changing NodeManager/ExprManager constructors to take Options
|
|
is not requested or eager type checking is enabled
|
|
|
|
|
|
|
|
|
|
|
|
* Add ContextMemoryAllocator<T> allocator type, conforming to
STL allocator requirements.
* Extend the CDList<> template to take an allocator (defaults
to std::allocator<T>).
* Add a specialized version of the CDList<> template (in
src/context/cdlist_context_memory.h) that allocates a list
in segments, in context memory.
* Add "forward" headers -- cdlist_forward.h, cdmap_forward.h,
and cdset_forward.h. Use these in public headers, and other
places where you don't need the full header (just the
forward-declaration). These types justify their own header
(instead of just forward-declaring yourself), because they
are complex templated types, with default template parameters,
specializations, etc.
* theory_engine.h no longer depends on individual theory headers.
(Instead it forward-declares Theory implementations.) This is
especially important now that theory .cpp files depend on
TheoryEngine (to implement Theory::getValue()). Previously,
any modification to any theory header file required *all*
theories, and the engine, to be completely rebuilt.
* Support memory cleanup for nontrivial CONSTANT kinds. This
resolves an issue with arithmetic where memory leaked for
each distinct Rational or Integer that was wrapped in a Node.
|
|
(resolves bug #212)
* also closed some other type checking loopholes in SmtEngine
* small fixes to define-sort (resolves bug #214)
* infrastructural support for printing expressions in languages
other than the internal representation language using an IO
manipulator, e.g.:
cout << Expr::setlanguage(language::output::LANG_SMTLIB_V2) << expr;
main() sets the output language for all streams to correspond to
the input language
* support delaying type checking in debug builds, so that one can debug
the type checker itself (before it was difficult, because debug builds did
all the type checking on Node creation!): new command-line flag
--no-early-type-checking (only makes sense for debug builds)
* disallowed copy-construction of ExprManager and NodeManager, and made other
constructors explicit; previously it was easy to unintentionally create
duplicate managers, with really weird results (i.e., disappearing
attributes!)
|
|
working (just need to decide where to expand)
|
|
modified in this commit
|
|
doesn't exist, and clean up a few things in NodeManager
|
|
also configure script auto-detection of __thread support and syntax
|
|
0 (resolves bug #200); on NodeManager/ExprManager side, no more prepareToBeDestroyed() / inDestruction
|
|
|
|
* src/theory/arith/normal_form.{h,cpp} contains the description for the new
normal form as well as utilities for dealing with the normal form.
* src/theory/arith/next_arith_rewriter.{h,cpp} contains the new rewriter.
The new rewriter implements preRewrite() and postRewrite() for arithmetic.
* src/theory/arith/arith_rewriter.{h,cpp} have been removed.
* TheoryArith::rewrite() has been removed.
* Arithmetic with the new normal form outperforms the trunk where the branch
occurred (-r797) on 46% of the examples in QF_LRA. (33% have no noticeable
difference.) Some important optimizations are stilling pending to the code
for handling the new normal form. (Bug 196.)
|
|
CongruenceClosure implementation; CongruenceClosure white-box test.
New UF theory implementation based on new CC module. This one
supports predicates. The two UF implementations exist in parallel
(they can be selected at runtime via the new command line option
"--uf").
Added type infrastructure for TUPLE.
Fixes to unit tests that failed in 16-August-2010 regressions.
Needed to instantiate TheoryEngine with an Options structure, and
explicitly call ->shutdown() on it before destruction (like the
SMTEngine does).
Fixed test makefiles to (1) perform all tests even in the presence of
failures, (2) give proper summaries of subdirectory tests
(e.g. regress0/uf and regress0/precedence)
Other minor changes.
|
|
|
|
* Added regression documentation to test/regress/README
* Added ability to print types of vars in expr printouts
with iomanipulator Node::printtypes(true)... for example,
Warning() << Node::printtypes(true) << n << std::endl;
* Types-printing can be specified on the command line with
--print-expr-types
* Improved type handling facilities and theoryOf().
For now, SORT_TYPE moved from builtin theory to UF theory
to match old behavior.
* Additional gdb debug functionality. Now we have:
debugPrintNode(Node) debugPrintRawNode(Node)
debugPrintTNode(TNode) debugPrintRawTNode(TNode)
debugPrintTypeNode(TypeNode) debugPrintRawTypeNode(TypeNode)
debugPrintNodeValue(NodeValue*) debugPrintRawNodeValue(NodeValue*)
they all print a {Node,TNode,NodeValue*} from the debugger.
The "Raw" versions print a very low-level AST-like form.
The regular versions do the same as operator<<, but force
full printing on (no depth-limiting).
* Other trivial fixes
|
|
* added TheoryArith::preRewrite() to test and demonstrate
the use of pre-rewriting.
* array types and type checking now supported
* array type checking now supported
* theoryOf() dispatching properly to arrays now
* theories now required to implement a (simple) identify()
function that returns a string identifying them for
debugging/user output purposes
* added "builtin" theory to hold all built-in kinds and their
type rules and rewriting (currently only exploding distinct)
* fixed production build failure (regarding NodeSetDepth)
* removed an errant "using namespace std" in util/bitvector.h
(and made associated trivial fixes elsewhere)
* fixes to make unexpected exceptions more verbose in debug builds
* fixes to make multiple, cascading assertion fails simpler
* minor other fixes to comments etc.
|
|
data into a CDMap. Such a key doesn't disappear from the map on pop,
but rather returns to its "initializing" state, set by
insertAtContextLevelZero(). This can be used for lazy assignment,
among other things, and has been added to support some exploratory
coding by Tim in arithmetic.
* Made internal CDOmap<> copy constructor private (it should always have
been). This is necessary to avoid CxxTest (or others) doing nasty
generic programming things that cause context invariants to be broken.
* Added unit testing for this feature, and in general beef up the unit
testing for CDMap<>.
* src/expr/node_manager.cpp: Better output for unhandled cases in getType().
|
|
|